Digital Processing

Peer-to-peer (P2P) botnets are the modern and most resilient bot structures which are harder to take down and stealthier to detect their malicious activities, because of which these are adopted by many of the recent botmasters. In this paper, we propose a novel botnet detection system which is capable to identify resilient P2P botnets. Our system initially identifies the p2p communication hosts present in the network. It then derives p2p traffic and further distinguishes between the botnet generated traffic and legitimate generated traffic. The parallelized computation makes scalability a default feature of our system. High detection accuracy and prodigious scalability are the extra features of our proposed system.

The goal of this project is to automatically detect and recognize some objects in an image by using a multi-agents architecture. A knowledge database is thus necessary and should contain an invariant description of known objects for the desired application. Point matching is an important aim in the research work. A digital image may undergo any arbitrary translational, rotational changes because of which the object shape may change. In this paper we present a novel approach for shape and feature matching using putative point matching.

This paper presents a proposal of detection of brain tumor of fetus through its MRI image during 25-27 week of pregnancy. This detection is based on edge detection technique and its implementation on FPGA. For FPGA implementation use MATLAB/SIMULINK graphical user interface with Xilinx System Generator which is the integrated design environment for FPGA within ISE11.3 design suit.

In parametric speech synthesis algorithms, a sequence of signals corresponding to phonemes is generated. However, synthesized speech tends to be unnatural as the vocal tract transition from one phoneme to another is not considered in most of the existing algorithms. This paper attempts to model the phoneme transition by extracting the speech parameters by means of Linear Time Varying system based Auto Regressive model. To reduce the capacity Speech parameters were represented in a polynomial equations. Sinusoidal Noise model was used to reconstruct the phoneme transition region. The results show moderate correlation of reconstructed transition regions with synthesized signal for different orders of polynomial.

Mobile video-audio transmission systems have delivered patient video with relevant snoring sound to quantify the severity of the sleep apnea patient over wireless networks, but few have optimized video-audio transmission in combination with transmission protocol over error-prone environments using wireless links. In this paper, the performance of the MPEG (Motion Picture Expert Group)-4 error resilient tools with UDP(User Datagram Protocol) protocol were evaluated over a wireless network to suggest the optimum combination of MPEG-4 error resilient tools and UDP packet size suitable for real-time transmission of video-audio transmission over error-prone mobile networks. Through experimentation, it was found that the packet size should correspond to IP(Internet Protocol) datagram size minus UDP and IP header for optimal video-audio quality. Also, for error resilient tool selection, the combination of resynchronization marker and data partitioning showed the best performance.

Automatic annotation is in fact the process of classifying medical using global and local features of standard image codes (IRMA) while being extracted. This includes four technical data axes of providing image (modality), direction, anatomy, and biological system. A number of recent researches have been conducted on the extraction of the scale invariant feature transform for automatic annotation, but until now no complete comparison has been conducted on the accuracy of the different classifications in resolution and annotation of the images based on the scale invariant feature transform. The results from the known and famous classifiers used on the four characteristics of anatomy, direction, biological system and modality are presented in this paper, which show that Sequential Minimal Optimization is the most efficient classification group as far as accuracy is concerned.

Automatic annotation is in fact the process of classifying medical using global and local features of standard image codes (IRMA) while being extracted. This includes four technical data axes of providing image (modality), direction, anatomy, and biological system. A few number of recent researches have been conducted on the extraction of Gabor filter feature in HSV color space for automatic annotation, but until now no complete comparison has been conducted on the accuracy of the different classifications in resolution and annotation of the images based on the Gabor filter feature in HSV color space. The results from the known and famous classifiers used on the four characteristics of anatomy, direction, biological system and modality are presented in this paper, which show that K-Nearest Neighbor is the most efficient classification group as far as accuracy is concerned.

Automatic license plate recognition of vehicles is a real time embedded system which identifies the characters directly from the image of the vehicle license plate. In the present research, an efficient method for license plate localization in the images without dependence of the plate color is proposed. After capturing the image of the vehicle, some image-enhancements are done first in order to reduce problems such as low quality and low contrast in the vehicle images. The proposed method extracts edges and then determines the candidate regions by comparing it with a selected plate which has appropriate size for selected capture distance. Finally by connected component elements analysis, the license plate is detected. The proposed system has improved efficiency as compared to earlier methods.

Image provides an important way for medical studies. The objective of the paper is to delineate the imaging process from graph theory with computer. The aim of digital image processing is to seek for the better methods and technologies on image process by noise riding, strengthening, restoring, dividing, extracting and so on based on computer. The main facts including the development of computer science, mathematics science and the increasing of application requirements on medical science, environmental science, industry and so on. Graph theory of computer provides a basic method and process for imaging process.

This paper proposed new techniques based on fuzzy logic in optimal direction for suppression of random valued impulse noise in digital images. The deviation of the test pixel from its neighboring pixels present in the optimal direction shows its corruption level. According to how much a pixel is impulse-like, a fuzzy index is assigned to each and every pixel in the image. After the detection of impulsivity of each pixel, a non-local mean filter is employed for noise suppression. Extensive simulations are performed to prove that the proposed techniques give better visual quality and quantitative measurement than recent impulse noise suppression techniques.